Episode Transcript
[00:00:00] Well, good morning, everybody. I just got back from Europe and it's a big week. There was also a very significant meeting over at Workday, disclosing their new product strategy to a lot of analysts, which I've gotten the briefing on. And I want to talk about enterprise AI and where we are. So I'm going to have a little bit of a long podcast this morning, but I think you guys will find it useful in my particular situation. I was in London with a bunch of companies and I was in Amsterdam with a bunch of companies and met up with a bunch of vendors there. The interesting thing about enterprise AI is the world is not evenly distributed. I talked to three very large banks, one of which is doing a very sophisticated agentic application for onboarding, partnering with ServiceNow, building out all sorts of use cases, reorganizing HR around it, quite sophisticated, very much in the direction of HR 2030. The second bank is a very large bank that's going through restructuring and they're doing what I consider to be the maybe not best solution, but they're going to go through all of the jobs in the bank and identify the tasks of those jobs and skills and then try to determine how to automate all the jobs they have. Only argument against that is that doesn't really allow you to redesign the processes. It just kind of gives people tools. We call this stage one, stage two, automation. And then the third bank is trying to figure out their new HR operating model and sort of starting from scratch and doesn't really have an AI strategy at all. So these are three very large companies, brand names that are at very, very different levels of maturity. So my point is that the world of enterprise AI is quite spiky, as they would say, in AI land. And some of you are going to be very sophisticated and some of you are not. And then I met with Sana and I met with a bunch of other companies over there and gave a keynote. And I won't replay that for you because you can probably find pieces of it around. But we're readying. We're. By the way, we're very close to launching the HR 2030 briefing paper, which is quite detailed. We'll probably launch it at our conference, if not before. So we have a lot of really good insights into where this is all going. But let me talk about Workday for a minute. While I was traveling, there was a very big meeting here in California for industry analysts and, and Workday kind of took the kimono, opened the kimono, as the expression goes, to explain where they're going with AI and with sauna and agents. And I was not there, but I got a lot of information from our team who was there. And I'm certainly going to keep working closely with Workday. The strategy is actually pretty cool and it makes sense from their standpoint. First of all, the Workday system, for those of you that have it, you know, and you may or may not like it, but you have it. You is filled with data, business rules, security rules, workflows that you've built around your company for the last 10 years. And the reason people bought Workday in the early days was to get off of the legacy systems you had that were either client, server or on prem into the cloud. And the purpose or the value of Workday was to build and implement and a system could be updated regularly in the cloud. You wouldn't fall behind, have all sorts of old integrations that were burdens for it. But what didn't happen with Workday is that it wasn't as flexible and dynamic as we thought. And they know that because it was designed in 2008. So, you know, we didn't have AI at the time. So people who have Workday are a little bit stuck that it's. I wouldn't call it a legacy system, but it's an inflexible system to some degree. And a lot of that gets back to the way the job architectures are implemented and the way we now think about dynamic work. But we didn't think about it that way 10 years ago. Many reasons for this. And so what Workday is doing with Sana is giving you a layer of easy to use, easy to implement, context aware AI software to leverage the infrastructure and business rules and data privacy that you have. And the argument they're kind of making, and they did a demo on this, is that if you just stick Claude or OpenAI or OpenClaw on top of your enterprise software, you're going to have all sorts of data problems, security problems and potentially business risk because you won't have leveraged all that infrastructure. And I know a lot of companies are going to buy that story because that's the way they think. And I just sat through three of them face to face or four of them in London. So they're pretty much saying we're the providers of the mainframe COBOL system. And just to use an analogy, I'm not saying it's exactly the same and we're going to make it easy for you to add on top of it without replacing it, because there really isn't a big alternative. If you, if you don't like Workday for some reason and you're tired of paying their fees, what are you going to do? Switch to SuccessFactor? Switch to Oracle? Hibob isn't big enough yet to replace Workday, and there aren't that many alternatives for a big global company. And so Workday is going to start charging you for access to the backend system in a different way than they do today. And so you may end up paying license fees to Workday in a different form. I'm not saying it's higher, it may not be higher than what you're paying now, but it may be in a different form. And then you're going to have all these tools from Sana and others, people like us, to lay on top of the system to build a new experience. And the nice thing about the story is you get to go as fast or slow as you want. You can use the off the shelf agents that they provide or, which is, you know, fine, a lot of those you might like and then. Or you can build stuff on top of it. And Sana is a very flexible, easy to use development environment. We know that. We've built a lot of stuff on it and it's just going to get better and better and better as they add more vibe coding tools on top of it. And you won't have to build all those business rules inside. Workday also understands that the magic of all of this AI enterprise stuff is the context layer. Not the actual agent itself, but the rules and the data elements behind it. The invisible stuff that you've been building all these years, the skills models, the career models, the workflows, the job architectures, the stuff that's buried in there that somebody designed for some particular purpose that isn't explicitly designed in the database, but sits on top of the database or around the database, the curricula in the lms. I mean, all of these things that companies build that are company specific, context awareness, data structures, we need to bring that with us into the world of agents. That's the argument they're making. You could start from scratch and over time these things will be redesigned in an AI form. But Workday is saying we're going to focus on that. And that's basically what we've been doing with Galileo. By the way, I won't give you the whole Galileo story, but, but we're moving Galileo into these other platforms for this precise reason, because context is everything in the world of AI, because the system does a lot of the workflow work that you use to program by hand. So you need to give it context so it can do what it's good at. So Arcdale understands that and that's great. It's taking them in the right direction. Now, if you think about the other providers in the world, I wrote a long article about Microsoft the other day and I would suggest you guys read it. Microsoft is also very much getting this.
[00:07:23] And what they did over the last six months or last. Last six, six weeks is, is they reorganized the copilot engineering into one unit. So now the Copilot is in a sense, a what's called a harness or a container for LLM models. Sort of like what SANA is, by the way. That's kind of what SANA is, really. It's a harness on top of the models. You can use multiple models. Well, that's what the copilot is. So for those of you that are big Microsoft shops and a lot of you I know are now have the Copilot, A lot of what Workday is talking about is the same thing that Microsoft's talking about now. Microsoft doesn't have the backend stuff. They're going to have to work with companies like Workday to do that, or third parties. But so this is kind of where the world is going. While all that's happening, of course, Anthropic and OpenAI are capturing our imaginations and showing us that maybe you don't need any of this. Just use my tool and it'll just whip up what you need and it's really attractive. I mentioned this to a bunch of people I was meeting with. We're redoing all of our websites right now. And the traditional way we've done this in the past, because I've done this at least 10 times, is you sit down with designers and you come up with pages and you come up with buttons and user experience interfaces and stuff, and then you kind of implement it. Well, you can do that in real time with these tools. So you don't have to mock up an application on the whiteboard. You can mock up an application in English and let Claude or OpenAI show show you what you just designed before you touch a piece of code. So when Workday or somebody else says, no, no, no, that's not a good idea, you got to use our stuff because it's safer and more secure. Some people are going to go, yeah, thank you for telling me that. I appreciate that. Some people are going to say, no, sorry, I'm just going to do it. Anyway, I'll figure it on my own. So there is an innovative new world emerging of people building stuff that won't necessarily leverage the more conservative approach that Workday is trying to get you to agree to. But I know most of you in the bigger companies are going to love this. And so a lot of what Anil is trying to do and Garrett and the people running Workday now is reinvent the company. In some sense. Workday is very rapidly being reinvented in the agentic world. A little bit along the way. IBM tried to do it back in the mainframe days. We don't need to use IBM as an example because they had lots of different diversions in the process. But I think it's really going to be attractive and a lot of you are going to like it. By the way, the other thing about the workday world is learning. Sana learning, or what we call Galileo learn. The other big to me, existential change in business is not just the agentic implementation of workflows and business processes and cross functional applications, but the fact that every employee gets can get information support, education data at their fingertips a hundred times faster than they could in the past. And what that means is that the ability for you to dynamically redeploy people, upskill people, engage people, develop people is spectacularly different in an agentic world than it was before. And I don't think Workdays is messaging that enough. I'll talk to them about it. But that is another, you know, significant part of what we're doing in hr. Moving to a world of dynamic enablement at all levels, not just in HR or in L and D, but really at every level of business. And we did meet. One of the dinners I had when I was in Europe is with a bunch of L and D people. And I sort of took them through the story for a good half hour and I'll tell you, the light bulbs went off and everybody is really thinking about the world differently. So there's a lot of opportunities here for you guys to focus. Now let's see. So what else happened? We also had a lot of discussions with vendors of different shapes and sizes, and I would sort of sympathize with all of you that unfortunately you're going to be flooded with demos and salespeople and marketing pitches from agent companies, even though many of them are interesting and a lot of them are going to be cool over time. There's two things I would encourage you to think about based on what's going on right this minute. Number one, as exciting as any agent tool may be, the ability for you to integrate it into your environment and create an architecture around it is going to be really fundamentally important, even if you're a small company, because the interoperability between these things is very immature. And nobody's going to be all workday or all sauna or all this or all that or even all Microsoft. So I won't get you into the details of this, but everybody's worried about that. Agent Sprawl, it's called. So that's number one. Number two is the fact that I don't think you're going to be successful working with a vendor that isn't rolling up their sleeves, helping you with agent design and implementation. You know, in the old world of innovative software companies, the tool was great and once everybody liked it, they mostly were a sales company and a product company. I don't think AI companies can be sales and product companies. They have to be partners with their customers in the enterprise world because this is so immature that what you do with it is likely to be something that they've never seen anybody do before. And the closer they are to you in understanding what you're doing. It's called forward engineering, I think it's called, in other words, giving you a technical resource to help you and work with you. If they're not able to do that or willing to do that, you might not want to work with that vendor. And that's a challenge for the vendors. A lot of them don't like those business models. The VCs don't like companies to spend money on services because the margins are as high. But I think the world is showing that the solution providers that work with you to build and implement and understand what your needs are are the ones that are going to thrive over time because it's just too immature. So if somebody's got some whiz bang AI thing and they just want you to buy it and they give you a bunch of references and they don't have somebody to help you really learn how to implement it and walk you through the process and get to know your business needs, I would just be a little bit careful because nothing's that mature yet. Okay. The other news I want to highlight this weekend is the layoffs at Meta, the early retirement program at Microsoft, the change in family leave and other benefits at Deloitte, and the surveillance that Facebook is doing of their employees. Now, there's a lot of news out there on stuff, so let me briefly just recap. Both Meta and Microsoft announced Layoffs and or early retirement policies. And as I've discussed extensively and had a lot of discussion about this in Europe, these are basically examples of tech companies who are going through internal business transformations. There's no question that it's easier to write code than it ever was before.
[00:14:21] So for coding, testing and coding, administration and code generation, there are definitely jobs probably that aren't needed anymore. But the general finding around the whole world of software engineering is that the number of software engineering jobs is actually flat to increasing. Because as more and more code can be created more quickly, we can build more systems and the systems are more complex. But you know, if you're Meta or Microsoft or another software company, you're trying to pivot to AI so fast, you're likely going to have individuals, teams and projects that are not aligned with the new direction. And so there's dislocation that's going to happen. And as I was saying to a reporter the other day, all of the software companies in the world are going through some form of internal crisis of how do we get from point A to point B? Point A being the traditional model of software where we develop it by hand and we sell a transactional solution of some kind to point B, which is we sell an AI agent. I think a lot of this dislocation is just that moving resources from project A to project B. That's certainly what's going on at Facebook Meta and some of it's rethinking the business model model of what people are working on. Because the interesting sort of existential change in software isn't just the technology that people are developing and how they're selling it and what they're selling, but how they're monetizing it. Because all of the business models for software in my life have been around user based pricing. X amount of dollars per month or per year per user. That's what you pay for Microsoft software, that's what you pay for Google, that's what you pay for all the corporate stuff. And then you size your business based on the number of users who buy it. Well, now we're in the electric utility business of selling capacity. So the new model is much more around how much activity and generative value are you actually generating and therefore how does your revenue flow through that lens. And of course, the corporations who buy the software don't have budgets that model that very well because we don't know how much we're going to use anything. There could be 10% of the users in a company that use 90% of the software because of the utility that they have. I was, you know, in a meeting, several meetings this week where software engineering teams discovered that the cost of using the generative AI was very, very high, higher than they thought. And maybe it actually would be cheaper to have a human do it who might be located in India at a lower cost of living location. So the software engineering industry, the product industry at least, is going through a lot of changes. So. So when you see layoffs at a software company, don't assume it's just the simple they automated away that job. It's much more complex than that. Now, when you read about Deloitte and other companies reducing benefits, that could just be cost cutting. But I think there's also a frustration that companies have about the number and volume and complexity of benefits that they've piled on to the employee base over the last years, from up until the pandemic and through the pandemic. And this mindset that we're going to give employees everything we possibly can to make them happy and to get them to come work for us kind of flies in the face of talent density. So I think those stories are, you know, interesting to read, but they're really kind of a reflection of the shift back towards productivity. And, you know, maybe we shouldn't pay people more money than they're worth unless they're highly productive, because in a talent density world, we have much more differentiated pay based on outcomes and just layering on or peanut buttering on lots and lots and lots and lots of benefits. It doesn't necessarily look like the right strategy when we're not constantly looking for new people to hire. So I think that's what that story's about. And then the final point I'll just talk about is surveillance. So there's a really interesting story about how Facebook is monitoring the keystrokes of their employees now and supposedly studying it for AI purposes. But you know, you can do anything with that data. And this story was written about by Jodi Cantor in the New York Times a year ago and you know, there's been a lot of maybe left wing, anti business opinions about this for a long time. Where I land on it just as an analyst, not as a lawyer, is when you go to work, the company is paying you for your efforts and energy and time to do the job you have.
[00:18:44] And if they are monitoring what you're doing, I think that's their right. The question is, what are they doing with the data? If they're using the data to make the work experience better, to Improve productivity, to better understand the systems that you're using, to make you a better employee and to make you more effective. To give you coaching, to give you development, give you advice if they're using it to surreptitiously measure your performance and secretly give you, giving you an invisible performance review that later results in you getting fired. That's not so good. It's not good for several reasons. First of all, it's probably illegal. I don't know what the law would be, but there's probably some legal issues there. I know there is in the EU and it's just bad business because human beings are not machines. They come to work with aspirations to succeed and they want to know. We all know, want to know, including me. What are we doing well, what could we be doing better? What are our weak points so we can improve and grow? And that level of energy and passion and ambition is a positive step in every company's evolution of growth. If you take that away and you turn it into a secret penalty or a secret performance evaluation, you create a sense of fear. And not only are you going to notice be able to hire people, but the people in your company are going to become very risk averse and they may work hard, but you're not going to get the best out of them. So I don't really know what's going on in Meta. It's a very interesting story and maybe they're using the data to train the AI on what the best practices are. And I think that's going to happen. I think any company where we have a lot of people doing similar jobs and we can look at what the high performers do and tell other people what they do so that they can learn how to be higher performers. I think that's fine. I think that's good. That's coaching and development, that's training and development, that's performance improvement. That's what the Toyota manufacturing process was all about, with people being willing and able to stop the manufacturing line when something's broken and tell everybody else what needs to be fixed. Those are quality projects. Those are good things to do. We can call it surveillance if we want to kind of scare people, but let's just face it, we're all being surveilled all the time. I mean, every time you go online and click a button on something on the Internet, some data is going to some advertising system to send you an ad. So I think we need to just kind of get over this and just make sure that the use cases and the legal ramifications of this are adopted, and we as leaders have policies to use the data for positive purposes. Anyway, you can read more about that. I'll put some links in the podcast. So it's been a fun week. I just want to give you guys all these things to think about over the weekend, and we'll talk more next week with some other cool stuff. Bye for now.
